[X86][APX] Remove KEYLOCKER and SHA promotions from EVEX MAP4. #89173
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@llvm/pr-subscribers-backend-x86 @llvm/pr-subscribers-mc Author: Freddy Ye (FreddyLeaf) Changes
Patch is 61.21 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/89173.diff 15 Files Affected:
diff --git a/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp b/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
index 4e4241efd63d6b..65e0c06a806e07 100644
--- a/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
+++ b/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
@@ -5046,18 +5046,14 @@ void X86DAGToDAGISel::Select(SDNode *Node) {
unsigned Opcode;
switch (IntNo) {
default: llvm_unreachable("Impossible intrinsic");
- case Intrinsic::x86_encodekey128:
- Opcode = GET_EGPR_IF_ENABLED(X86::ENCODEKEY128);
- break;
- case Intrinsic::x86_encodekey256:
- Opcode = GET_EGPR_IF_ENABLED(X86::ENCODEKEY256);
- break;
+ case Intrinsic::x86_encodekey128: Opcode = X86::ENCODEKEY128; break;
+ case Intrinsic::x86_encodekey256: Opcode = X86::ENCODEKEY256; break;
}
SDValue Chain = Node->getOperand(0);
Chain = CurDAG->getCopyToReg(Chain, dl, X86::XMM0, Node->getOperand(3),
SDValue());
- if (Opcode == X86::ENCODEKEY256 || Opcode == X86::ENCODEKEY256_EVEX)
+ if (Opcode == X86::ENCODEKEY256)
Chain = CurDAG->getCopyToReg(Chain, dl, X86::XMM1, Node->getOperand(4),
Chain.getValue(1));
@@ -6476,18 +6472,17 @@ void X86DAGToDAGISel::Select(SDNode *Node) {
default:
llvm_unreachable("Unexpected opcode!");
case X86ISD::AESENCWIDE128KL:
- Opcode = GET_EGPR_IF_ENABLED(X86::AESENCWIDE128KL);
+ Opcode = X86::AESENCWIDE128KL;
break;
case X86ISD::AESDECWIDE128KL:
- Opcode = GET_EGPR_IF_ENABLED(X86::AESDECWIDE128KL);
+ Opcode = X86::AESDECWIDE128KL;
break;
case X86ISD::AESENCWIDE256KL:
- Opcode = GET_EGPR_IF_ENABLED(X86::AESENCWIDE256KL);
+ Opcode = X86::AESENCWIDE256KL;
break;
case X86ISD::AESDECWIDE256KL:
- Opcode = GET_EGPR_IF_ENABLED(X86::AESDECWIDE256KL);
+ Opcode = X86::AESDECWIDE256KL;
break;
-#undef GET_EGPR_IF_ENABLED
}
SDValue Chain = Node->getOperand(0);
diff --git a/llvm/lib/Target/X86/X86InstrAsmAlias.td b/llvm/lib/Target/X86/X86InstrAsmAlias.td
index d06a0c79b46bbb..25b8297502f60d 100644
--- a/llvm/lib/Target/X86/X86InstrAsmAlias.td
+++ b/llvm/lib/Target/X86/X86InstrAsmAlias.td
@@ -894,11 +894,3 @@ def : InstAlias<"vmsave\t{%rax|rax}", (VMSAVE64), 0>, Requires<[In64BitMode]>;
def : InstAlias<"invlpga\t{%eax, %ecx|eax, ecx}", (INVLPGA32), 0>, Requires<[Not64BitMode]>;
def : InstAlias<"invlpga\t{%rax, %ecx|rax, ecx}", (INVLPGA64), 0>, Requires<[In64BitMode]>;
-// Aliases with explicit %xmm0
-def : InstAlias<"sha256rnds2\t{$src2, $dst|$dst, $src2}",
- (SHA256RNDS2rr VR128:$dst, VR128:$src2), 0>;
-def : InstAlias<"sha256rnds2\t{$src2, $dst|$dst, $src2}",
- (SHA256RNDS2rm VR128:$dst, i128mem:$src2), 0>;
-
-def : InstAlias<"sha256rnds2\t{$src2, $dst|$dst, $src2}",
- (SHA256RNDS2rm_EVEX VR128:$dst, i128mem:$src2), 0>;
diff --git a/llvm/lib/Target/X86/X86InstrKL.td b/llvm/lib/Target/X86/X86InstrKL.td
index 32423e08ed4c53..d5a4be45cc9b10 100644
--- a/llvm/lib/Target/X86/X86InstrKL.td
+++ b/llvm/lib/Target/X86/X86InstrKL.td
@@ -48,7 +48,7 @@ let SchedRW = [WriteSystem] in {
[(int_x86_loadiwkey XMM0, VR128:$src1, VR128:$src2, EAX)]>, T8, XS;
}
- let Predicates = [HasKL, NoEGPR] in {
+ let Predicates = [HasKL] in {
let Uses = [XMM0], Defs = [XMM0, XMM1, XMM2, XMM4, XMM5, XMM6, EFLAGS] in
def ENCODEKEY128 : Encodekey<0xFA, "encodekey128">, T8;
@@ -58,17 +58,6 @@ let SchedRW = [WriteSystem] in {
let Constraints = "$src1 = $dst", Defs = [EFLAGS] in
defm "" : Aesencdec<"">, T8;
}
-
- let Predicates = [HasKL, HasEGPR, In64BitMode] in {
- let Uses = [XMM0], Defs = [XMM0, XMM1, XMM2, XMM4, XMM5, XMM6, EFLAGS] in
- def ENCODEKEY128_EVEX : Encodekey<0xDA, "encodekey128">, EVEX, T_MAP4;
-
- let Uses = [XMM0, XMM1], Defs = [XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, EFLAGS] in
- def ENCODEKEY256_EVEX : Encodekey<0xDB, "encodekey256">, EVEX, T_MAP4;
-
- let Constraints = "$src1 = $dst", Defs = [EFLAGS] in
- defm "" : Aesencdec<"_EVEX">, EVEX, T_MAP4;
- }
} // SchedRW
multiclass Aesencdecwide<string suffix> {
@@ -80,9 +69,6 @@ multiclass Aesencdecwide<string suffix> {
let SchedRW = [WriteSystem], Uses = [XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7],
Defs = [EFLAGS, XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7], mayLoad = 1 in {
- let Predicates = [HasWIDEKL, NoEGPR] in
+ let Predicates = [HasWIDEKL] in
defm "" : Aesencdecwide<"">, T8;
-
- let Predicates = [HasWIDEKL, HasEGPR, In64BitMode] in
- defm "" : Aesencdecwide<"_EVEX">, EVEX, T_MAP4;
} // SchedRW
diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td
index 2b391b60f2c9bb..8542e5ce233364 100644
--- a/llvm/lib/Target/X86/X86InstrSSE.td
+++ b/llvm/lib/Target/X86/X86InstrSSE.td
@@ -6689,31 +6689,31 @@ let Predicates = [HasCRC32, HasEGPR, In64BitMode], OpMap = T_MAP4, OpEnc = EncEV
// FIXME: Is there a better scheduler class for SHA than WriteVecIMul?
multiclass SHAI_binop<bits<8> Opc, string OpcodeStr, Intrinsic IntId,
- X86FoldableSchedWrite sched, string Suffix = "", bit UsesXMM0 = 0> {
- def rr#Suffix : I<Opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
- !if(UsesXMM0,
- !strconcat(OpcodeStr, "\t{%xmm0, $src2, $dst|$dst, $src2, xmm0}"),
- !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}")),
- [!if(UsesXMM0,
- (set VR128:$dst, (IntId VR128:$src1, VR128:$src2, XMM0)),
- (set VR128:$dst, (IntId VR128:$src1, VR128:$src2)))]>,
- T8, Sched<[sched]>;
-
- def rm#Suffix : I<Opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, i128mem:$src2),
- !if(UsesXMM0,
- !strconcat(OpcodeStr, "\t{%xmm0, $src2, $dst|$dst, $src2, xmm0}"),
- !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}")),
- [!if(UsesXMM0,
- (set VR128:$dst, (IntId VR128:$src1,
- (memop addr:$src2), XMM0)),
- (set VR128:$dst, (IntId VR128:$src1,
- (memop addr:$src2))))]>, T8,
- Sched<[sched.Folded, sched.ReadAfterFold]>;
+ X86FoldableSchedWrite sched, bit UsesXMM0 = 0> {
+ def rr : I<Opc, MRMSrcReg, (outs VR128:$dst),
+ (ins VR128:$src1, VR128:$src2),
+ !if(UsesXMM0,
+ !strconcat(OpcodeStr, "\t{%xmm0, $src2, $dst|$dst, $src2, xmm0}"),
+ !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}")),
+ [!if(UsesXMM0,
+ (set VR128:$dst, (IntId VR128:$src1, VR128:$src2, XMM0)),
+ (set VR128:$dst, (IntId VR128:$src1, VR128:$src2)))]>,
+ T8, Sched<[sched]>;
+
+ def rm : I<Opc, MRMSrcMem, (outs VR128:$dst),
+ (ins VR128:$src1, i128mem:$src2),
+ !if(UsesXMM0,
+ !strconcat(OpcodeStr, "\t{%xmm0, $src2, $dst|$dst, $src2, xmm0}"),
+ !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}")),
+ [!if(UsesXMM0,
+ (set VR128:$dst, (IntId VR128:$src1,
+ (memop addr:$src2), XMM0)),
+ (set VR128:$dst, (IntId VR128:$src1,
+ (memop addr:$src2))))]>, T8,
+ Sched<[sched.Folded, sched.ReadAfterFold]>;
}
-let Constraints = "$src1 = $dst", Predicates = [HasSHA, NoEGPR] in {
+let Constraints = "$src1 = $dst", Predicates = [HasSHA] in {
def SHA1RNDS4rri : Ii8<0xCC, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, u8imm:$src3),
"sha1rnds4\t{$src3, $src2, $dst|$dst, $src2, $src3}",
@@ -6740,7 +6740,7 @@ let Constraints = "$src1 = $dst", Predicates = [HasSHA, NoEGPR] in {
let Uses=[XMM0] in
defm SHA256RNDS2 : SHAI_binop<0xCB, "sha256rnds2", int_x86_sha256rnds2,
- SchedWriteVecIMul.XMM, "", 1>;
+ SchedWriteVecIMul.XMM, 1>;
defm SHA256MSG1 : SHAI_binop<0xCC, "sha256msg1", int_x86_sha256msg1,
SchedWriteVecIMul.XMM>;
@@ -6748,47 +6748,11 @@ let Constraints = "$src1 = $dst", Predicates = [HasSHA, NoEGPR] in {
SchedWriteVecIMul.XMM>;
}
-let Constraints = "$src1 = $dst", Predicates = [HasSHA, HasEGPR, In64BitMode] in {
- def SHA1RNDS4rri_EVEX: Ii8<0xD4, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2, u8imm:$src3),
- "sha1rnds4\t{$src3, $src2, $dst|$dst, $src2, $src3}",
- [(set VR128:$dst,
- (int_x86_sha1rnds4 VR128:$src1, VR128:$src2,
- (i8 timm:$src3)))]>,
- EVEX, NoCD8, T_MAP4, Sched<[SchedWriteVecIMul.XMM]>;
- def SHA1RNDS4rmi_EVEX: Ii8<0xD4, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, i128mem:$src2, u8imm:$src3),
- "sha1rnds4\t{$src3, $src2, $dst|$dst, $src2, $src3}",
- [(set VR128:$dst,
- (int_x86_sha1rnds4 VR128:$src1,
- (memop addr:$src2),
- (i8 timm:$src3)))]>,
- EVEX, NoCD8, T_MAP4,
- Sched<[SchedWriteVecIMul.XMM.Folded,
- SchedWriteVecIMul.XMM.ReadAfterFold]>;
-
- defm SHA1NEXTE : SHAI_binop<0xD8, "sha1nexte", int_x86_sha1nexte,
- SchedWriteVecIMul.XMM, "_EVEX">,
- EVEX, NoCD8, T_MAP4;
- defm SHA1MSG1 : SHAI_binop<0xD9, "sha1msg1", int_x86_sha1msg1,
- SchedWriteVecIMul.XMM, "_EVEX">,
- EVEX, NoCD8, T_MAP4;
- defm SHA1MSG2 : SHAI_binop<0xDA, "sha1msg2", int_x86_sha1msg2,
- SchedWriteVecIMul.XMM, "_EVEX">,
- EVEX, NoCD8, T_MAP4;
-
- let Uses=[XMM0] in
- defm SHA256RNDS2 : SHAI_binop<0xDB, "sha256rnds2", int_x86_sha256rnds2,
- SchedWriteVecIMul.XMM, "_EVEX", 1>,
- EVEX, NoCD8, T_MAP4;
-
- defm SHA256MSG1 : SHAI_binop<0xDC, "sha256msg1", int_x86_sha256msg1,
- SchedWriteVecIMul.XMM, "_EVEX">,
- EVEX, NoCD8, T_MAP4;
- defm SHA256MSG2 : SHAI_binop<0xDD, "sha256msg2", int_x86_sha256msg2,
- SchedWriteVecIMul.XMM, "_EVEX">,
- EVEX, NoCD8, T_MAP4;
-}
+// Aliases with explicit %xmm0
+def : InstAlias<"sha256rnds2\t{$src2, $dst|$dst, $src2}",
+ (SHA256RNDS2rr VR128:$dst, VR128:$src2), 0>;
+def : InstAlias<"sha256rnds2\t{$src2, $dst|$dst, $src2}",
+ (SHA256RNDS2rm VR128:$dst, i128mem:$src2), 0>;
//===----------------------------------------------------------------------===//
// AES-NI Instructions
diff --git a/llvm/test/CodeGen/X86/keylocker-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/keylocker-intrinsics-fast-isel.ll
index f91c4ad0e8e343..5ae1b06b072ff3 100644
--- a/llvm/test/CodeGen/X86/keylocker-intrinsics-fast-isel.ll
+++ b/llvm/test/CodeGen/X86/keylocker-intrinsics-fast-isel.ll
@@ -1,6 +1,5 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -fast-isel -mtriple=x86_64-unknown-unknown -mattr=+kl,+widekl | FileCheck %s
-; RUN: llc < %s -fast-isel -mtriple=x86_64-unknown-unknown -mattr=+kl,+widekl,+egpr --show-mc-encoding | FileCheck %s --check-prefix=EGPR
; NOTE: This should use IR equivalent to what is generated by clang/test/CodeGen/X86/keylocker-builtins.c
@@ -10,12 +9,6 @@ define void @test_loadiwkey(i32 %ctl, <2 x i64> %intkey, <2 x i64> %enkey_lo, <2
; CHECK-NEXT: movl %edi, %eax
; CHECK-NEXT: loadiwkey %xmm2, %xmm1
; CHECK-NEXT: retq
-;
-; EGPR-LABEL: test_loadiwkey:
-; EGPR: # %bb.0: # %entry
-; EGPR-NEXT: movl %edi, %eax # encoding: [0x89,0xf8]
-; EGPR-NEXT: loadiwkey %xmm2, %xmm1 # encoding: [0xf3,0x0f,0x38,0xdc,0xca]
-; EGPR-NEXT: retq # encoding: [0xc3]
entry:
tail call void @llvm.x86.loadiwkey(<2 x i64> %intkey, <2 x i64> %enkey_lo, <2 x i64> %enkey_hi, i32 %ctl)
ret void
@@ -32,17 +25,6 @@ define i32 @test_encodekey128_u32(i32 %htype, <2 x i64> %key, ptr nocapture %h)
; CHECK-NEXT: movups %xmm5, 64(%rsi)
; CHECK-NEXT: movups %xmm6, 80(%rsi)
; CHECK-NEXT: retq
-;
-; EGPR-LABEL: test_encodekey128_u32:
-; EGPR: # %bb.0: # %entry
-; EGPR-NEXT: encodekey128 %edi, %eax # EVEX TO LEGACY Compression encoding: [0xf3,0x0f,0x38,0xfa,0xc7]
-; EGPR-NEXT: movups %xmm0, (%rsi) # encoding: [0x0f,0x11,0x06]
-; EGPR-NEXT: movups %xmm1, 16(%rsi) # encoding: [0x0f,0x11,0x4e,0x10]
-; EGPR-NEXT: movups %xmm2, 32(%rsi) # encoding: [0x0f,0x11,0x56,0x20]
-; EGPR-NEXT: movups %xmm4, 48(%rsi) # encoding: [0x0f,0x11,0x66,0x30]
-; EGPR-NEXT: movups %xmm5, 64(%rsi) # encoding: [0x0f,0x11,0x6e,0x40]
-; EGPR-NEXT: movups %xmm6, 80(%rsi) # encoding: [0x0f,0x11,0x76,0x50]
-; EGPR-NEXT: retq # encoding: [0xc3]
entry:
%0 = tail call { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.x86.encodekey128(i32 %htype, <2 x i64> %key)
%1 = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %0, 1
@@ -78,18 +60,6 @@ define i32 @test_encodekey256_u32(i32 %htype, <2 x i64> %key_lo, <2 x i64> %key_
; CHECK-NEXT: movups %xmm5, 80(%rsi)
; CHECK-NEXT: movups %xmm6, 96(%rsi)
; CHECK-NEXT: retq
-;
-; EGPR-LABEL: test_encodekey256_u32:
-; EGPR: # %bb.0: # %entry
-; EGPR-NEXT: encodekey256 %edi, %eax # EVEX TO LEGACY Compression encoding: [0xf3,0x0f,0x38,0xfb,0xc7]
-; EGPR-NEXT: movups %xmm0, (%rsi) # encoding: [0x0f,0x11,0x06]
-; EGPR-NEXT: movups %xmm1, 16(%rsi) # encoding: [0x0f,0x11,0x4e,0x10]
-; EGPR-NEXT: movups %xmm2, 32(%rsi) # encoding: [0x0f,0x11,0x56,0x20]
-; EGPR-NEXT: movups %xmm3, 48(%rsi) # encoding: [0x0f,0x11,0x5e,0x30]
-; EGPR-NEXT: movups %xmm4, 64(%rsi) # encoding: [0x0f,0x11,0x66,0x40]
-; EGPR-NEXT: movups %xmm5, 80(%rsi) # encoding: [0x0f,0x11,0x6e,0x50]
-; EGPR-NEXT: movups %xmm6, 96(%rsi) # encoding: [0x0f,0x11,0x76,0x60]
-; EGPR-NEXT: retq # encoding: [0xc3]
entry:
%0 = tail call { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.x86.encodekey256(i32 %htype, <2 x i64> %key_lo, <2 x i64> %key_hi)
%1 = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %0, 1
@@ -123,13 +93,6 @@ define zeroext i8 @test_mm_aesenc256kl_u8(ptr %odata, <2 x i64> %idata, ptr %h)
; CHECK-NEXT: sete %al
; CHECK-NEXT: movaps %xmm0, (%rdi)
; CHECK-NEXT: retq
-;
-; EGPR-LABEL: test_mm_aesenc256kl_u8:
-; EGPR: # %bb.0: # %entry
-; EGPR-NEXT: aesenc256kl (%rsi), %xmm0 # EVEX TO LEGACY Compression encoding: [0xf3,0x0f,0x38,0xde,0x06]
-; EGPR-NEXT: sete %al # encoding: [0x0f,0x94,0xc0]
-; EGPR-NEXT: movaps %xmm0, (%rdi) # encoding: [0x0f,0x29,0x07]
-; EGPR-NEXT: retq # encoding: [0xc3]
entry:
%0 = tail call { i8, <2 x i64> } @llvm.x86.aesenc256kl(<2 x i64> %idata, ptr %h) #1
%1 = extractvalue { i8, <2 x i64> } %0, 1
@@ -145,13 +108,6 @@ define zeroext i8 @test_mm_aesdec256kl_u8(ptr %odata, <2 x i64> %idata, ptr %h)
; CHECK-NEXT: sete %al
; CHECK-NEXT: movaps %xmm0, (%rdi)
; CHECK-NEXT: retq
-;
-; EGPR-LABEL: test_mm_aesdec256kl_u8:
-; EGPR: # %bb.0: # %entry
-; EGPR-NEXT: aesdec256kl (%rsi), %xmm0 # EVEX TO LEGACY Compression encoding: [0xf3,0x0f,0x38,0xdf,0x06]
-; EGPR-NEXT: sete %al # encoding: [0x0f,0x94,0xc0]
-; EGPR-NEXT: movaps %xmm0, (%rdi) # encoding: [0x0f,0x29,0x07]
-; EGPR-NEXT: retq # encoding: [0xc3]
entry:
%0 = tail call { i8, <2 x i64> } @llvm.x86.aesdec256kl(<2 x i64> %idata, ptr %h) #1
%1 = extractvalue { i8, <2 x i64> } %0, 1
@@ -167,13 +123,6 @@ define zeroext i8 @test_mm_aesenc128kl_u8(ptr %odata, <2 x i64> %idata, ptr %h)
; CHECK-NEXT: sete %al
; CHECK-NEXT: movaps %xmm0, (%rdi)
; CHECK-NEXT: retq
-;
-; EGPR-LABEL: test_mm_aesenc128kl_u8:
-; EGPR: # %bb.0: # %entry
-; EGPR-NEXT: aesenc128kl (%rsi), %xmm0 # EVEX TO LEGACY Compression encoding: [0xf3,0x0f,0x38,0xdc,0x06]
-; EGPR-NEXT: sete %al # encoding: [0x0f,0x94,0xc0]
-; EGPR-NEXT: movaps %xmm0, (%rdi) # encoding: [0x0f,0x29,0x07]
-; EGPR-NEXT: retq # encoding: [0xc3]
entry:
%0 = tail call { i8, <2 x i64> } @llvm.x86.aesenc128kl(<2 x i64> %idata, ptr %h) #1
%1 = extractvalue { i8, <2 x i64> } %0, 1
@@ -189,13 +138,6 @@ define zeroext i8 @test_mm_aesdec128kl_u8(ptr %odata, <2 x i64> %idata, ptr %h)
; CHECK-NEXT: sete %al
; CHECK-NEXT: movaps %xmm0, (%rdi)
; CHECK-NEXT: retq
-;
-; EGPR-LABEL: test_mm_aesdec128kl_u8:
-; EGPR: # %bb.0: # %entry
-; EGPR-NEXT: aesdec128kl (%rsi), %xmm0 # EVEX TO LEGACY Compression encoding: [0xf3,0x0f,0x38,0xdd,0x06]
-; EGPR-NEXT: sete %al # encoding: [0x0f,0x94,0xc0]
-; EGPR-NEXT: movaps %xmm0, (%rdi) # encoding: [0x0f,0x29,0x07]
-; EGPR-NEXT: retq # encoding: [0xc3]
entry:
%0 = tail call { i8, <2 x i64> } @llvm.x86.aesdec128kl(<2 x i64> %idata, ptr %h) #1
%1 = extractvalue { i8, <2 x i64> } %0, 1
@@ -226,28 +168,6 @@ define zeroext i8 @test__mm_aesencwide128kl_u8(ptr %odata, ptr %idata, ptr %h) {
; CHECK-NEXT: movaps %xmm6, 96(%rdi)
; CHECK-NEXT: movaps %xmm7, 112(%rdi)
; CHECK-NEXT: retq
-;
-; EGPR-LABEL: test__mm_aesencwide128kl_u8:
-; EGPR: # %bb.0: # %entry
-; EGPR-NEXT: movaps (%rsi), %xmm0 # encoding: [0x0f,0x28,0x06]
-; EGPR-NEXT: movaps 16(%rsi), %xmm1 # encoding: [0x0f,0x28,0x4e,0x10]
-; EGPR-NEXT: movaps 32(%rsi), %xmm2 # encoding: [0x0f,0x28,0x56,0x20]
-; EGPR-NEXT: movaps 48(%rsi), %xmm3 # encoding: [0x0f,0x28,0x5e,0x30]
-; EGPR-NEXT: movaps 64(%rsi), %xmm4 # encoding: [0x0f,0x28,0x66,0x40]
-; EGPR-NEXT: movaps 80(%rsi), %xmm5 # encoding: [0x0f,0x28,0x6e,0x50]
-; EGPR-NEXT: movaps 96(%rsi), %xmm6 # encoding: [0x0f,0x28,0x76,0x60]
-; EGPR-NEXT: movaps 112(%rsi), %xmm7 # encoding: [0x0f,0x28,0x7e,0x70]
-; EGPR-NEXT: aesencwide128kl (%rdx) # EVEX TO LEGACY Compression encoding: [0xf3,0x0f,0x38,0xd8,0x02]
-; EGPR-NEXT: sete %al # encoding: [0x0f,0x94,0xc0]
-; EGPR-NEXT: movaps %xmm0, (%rdi) # encoding: [0x0f,0x29,0x07]
-; EGPR-NEXT: movaps %xmm1, 16(%rdi) # encoding: [0x0f,0x29,0x4f,0x10]
-; EGPR-NEXT: movaps %xmm2, 32(%rdi) # encoding: [0x0f,0x29,0x57,0x20]
-; EGPR-NEXT: movaps %xmm3, 48(%rdi) # encoding: [0x0f,0x29,0x5f,0x30]
-; EGPR-NEXT: movaps %xmm4, 64(%rdi) # encoding: [0x0f,0x29,0x67,0x40]
-; EGPR-NEXT: movaps %xmm5, 80(%rdi) # encoding: [0x0f,0x29,0x6f,0x50]
-; EGPR-NEXT: movaps %xmm6, 96(%rdi) # encoding: [0x0f,0x29,0x77,0x60]
-; EGPR-NEXT: movaps %xmm7, 112(%rdi) # encoding: [0x0f,0x29,0x7f,0x70]
-; EGPR-NEXT: retq # encoding: [0xc3]
entry:
%0 = load <2 x i64>, ptr %idata, align 16
%1 = getelementptr <2 x i64>, ptr %idata, i64 1
@@ -314,28 +234,6 @@ define zeroext i8 @test__mm_aesdecwide128kl_u8(ptr %odata, ptr %idata, ptr %h) {
; CHECK-NEXT: movaps %xmm6, 96(%rdi)
; CHECK-NEXT: movaps %xmm7, 112(%rdi)
; CHECK-NEXT: retq
-;
-; EGPR-LABEL: test__mm_aesdecwide128kl_u8:
-; EGPR: # %bb.0: # %entry
-; EGPR-NEXT: movaps (%rsi), %xmm0 # encoding: [0x0f,0x28,0x06]
-; EGPR-NEXT: movaps 16(%rsi), %xmm1 # encoding: [0x0f,0x28,0x4e,0x10]
-; EGPR-NEXT: movaps 32(%rsi), %xmm2 # encoding: [0x0f,0x28,0x56,0x20]
-; EGPR-NEXT: movaps 48(%rsi), %xmm3 # encoding: [0x0f,0x28,0x5e,0x30]
-; EGPR-NEXT: movaps 64(%rsi), %xmm4 # encoding: [0x0f,0x28,0x66,0x40]
-; EGPR-NEXT: movaps 80(%rsi), %xmm5 # encoding: [0x0f,0x28,0x6e,0x50]
-; EGPR-NEXT: movaps 96(%rsi), %xmm6 # encoding: [0x0f,0x28,0x76,0x60]
-; EGPR-NEXT: movaps 112(%rsi), %xmm7 # encoding: [0x0f,0x28,0x7e,0x70]
-; EGPR-NEXT: aesdecwide128kl (%rdx) # EVEX TO LEGACY Compression encoding: [0xf3,0x0f,0x38,0xd8,0x0a]
-; EGPR-NEXT: sete %al # encoding: [0x0f,0x94,0xc0]
-; EGPR-NEXT: movaps %xmm0, (%rdi) # encoding: [0x0f,0x...
[truncated]
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
KanRobert
reviewed
Apr 18, 2024
|
Could you list the opcodes for your change in the description? I noticed AESENCWIDE256KL is removed in your patch, is it expected? |
Done. it is expected. See version 4.0 |
KanRobert
reviewed
Apr 19, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
APX spec: https://cdrdv2.intel.com/v1/dl/getContent/784266
Change happended in version 4.0.
Removed instructions' Opcodes:
AESDEC128KL
AESDEC256KL
AESDECWIDE128KL
AESDECWIDE256KL
AESENC128KL
AESENC256KL
AESENCWIDE128KL
AESENCWIDE256KL
ENCODEKEY128
ENCODEKEY256
SHA1MSG1
SHA1MSG2
SHA1NEXTE
SHA1RNDS4
SHA256MSG1
SHA256MSG2
SHA256RNDS2